Environment for Model Maintenance, Integration and Tuning (EMMIT)

The EMMIT software platform provides a framework for hosting, integration, testing, and monitoring of machine learning models in a healthcare setting.

Value Proposition

Integrating machine learning models and predictive analytics with a hospital environment is difficult. Data flowing in and out of the model can become a black box with no way to monitor if the models are operating or data transfer is effective. EMMIT is a platform for hosting any number of unique models (or even multiple versions of the same model) and helping to deploy them within an EHR system (i.e., Epic, Cerner). It then provides operational monitoring (uptime, alerts, error reporting) to teams who would otherwise have no insight on the status of their software.

Associated Products

The FDA has declared an urgent need for improved model performance monitoring in AI/ML-based health analytics. The Model Monitoring Module is an add-on to EMMIT that aims to detect and alert to declines in model performance and drift.

Competitive Advantage

  • Provides access to information on data flowing into/out of the models

  • Improves model uptime with monitoring and alerts

  • Enables prospective analysis for researchers

  • Facilitates simplified, real-time A/B testing for model comparisons

Unique Features

  • Dashboard displays operational status: if all points of integration are working as expected or returning performance errors

  • Local or cloud-based operation

  • Hosts multiple models simultaneously

  • Modular add-ons in development to address future regulatory and maintenance requirements

Emerging Product These products have completed extensive research and development and are actively seeking licensing partners to take them to market.

Project Leads
Sardar Ansari, PhD
Joseph Blackmer

Licensing Manager
Drew Bennett

Intellectual Property
Invention Disclosure # 2022-108

Image credit: Shutterstock

INDUSTRY OPPORTUNITY
Typical model integration with 3rd party EHRs removes the ability to see the flow of data going in and out, leaving teams in the dark when it comes to the maintenance of their analytics. EMMIT removes the black-box style integration for models and analytics that work within Epic and other EHR systems by delivering operational monitoring dashboards to ensure systems are working as expected. Please contact the Licensing Manager, Drew Bennett, for more information.

Funding History

$121,563 in non-dilutive funding

  • $121,563 2022 MTRAC Grant

  • Substantial (additional) departmental, school and center based support

Completed milestones

  • Architecture and Design

  • Prototype development and testing

  • Operational Monitoring and visualization through Grafana

  • U of M integration with MiChart, full integration with U of M Infrastructure

Next Steps

  • Clinical trial

  • Real-time data pipeline leveraging the PACS system, HL7 and waveform data

  • Build-out of front end and permissions-based user levels

Funding Organizations

Publications

None at this time

Media Coverage

None at this time

Kate MurphyEMMETT, Predictive Analysis, AirStrip, MTRAC, EMMIT, model monitoring